I understand. Some more points.
1. If the laser is only illuminating one point at a time, the detector can simply be pointed at the whole specimen - no mirrors needed.
2. Rotating is inherently non-linear. The beam will be moving slowest at 90 degrees to the mirror and fastest at smaller angles. Move a led pointer from the straight up to 45 degrees along a wall.
3. Do you really need mirrors at all? Perhaps one slow scan will do it. Can the specimen me put on a stage with two servos, one for X and one for Y?
The your arduino moves the stage, turns on the led, reads the photodetector, turns off the led and repeats.
It's then all perfectly linear and controlled. Small 4 mm actuators are $10 ( http://www.sparkfun.com/products/8782
but larger, 50 and 100mm ones are ~ $100. ( http://www.trossenrobotics.com/linear-actuators.aspx
Also this way you can go slower and add up a lot of samples for a higher dynamic range of image.
Alternatively you can have the arduino just time how long the laser has to be on any given x y spot in order to reach a certain accumulated threshold of fluorescence.
Highly fluorescent samples would scan quickly while you'd be able to scan even slightly fluorescent ones, only much more slowly.
Oh if you can't move the rock, perhaps you could move the LED/detector part, but also with an x y stage for repeatability and linearity.
Maybe a good mechanical designer (not me) could do it with two inexpensive rotary servo motors.
What do you think?